It's been a while since I read it, but his experimental evidence was a lot of tiny studies with no replication. I tend to not view a single study as strong evidence at all, especially in the social sciences. And history hasn't been particularly kind to many of those findings.

I went in really wanting to love the book, it was universally praised and I was pretty high on Ariely at the time, but even with that bias I still found large chunks quite unconvincing.

Perhaps it was just the writing style or something, or the fact that the beginning of the book focused a lot on those priming studies at a time when that entire area of research was profoundly embarrassing the social sciences.

(re: Gladwell, that's a good way of putting it, it was uncomfortably Gladwellian! Not that level obviously, but not far enough away!)

Reading the book, I had to stop many times to reflect on how poorly constructed the conclusions were. Many of his conclusions suffered from the same biases that he had "discovered", and yet he appeared to be blithely unaware of his own biases.

My favorite part was his discussion of the Linda problem, where he would get different results depending on how the question was phrased. In the book, he never once considered that the problem may be that his formulations of the problem were not well communicated to the study participants.

In the end, I decided that his research into cognitive bias was more a journey of self exploration, but somehow he discovered these biases in everyone except himself.

> In the end, I decided that his research into cognitive bias was more a journey of self exploration, but somehow he discovered these biases in everyone except himself.

Maybe fixing humans is impossible and it would be better for meta-meta-cognition to be accomplished by other people. Clearly they are quite good at it already!

Suppose you had a complete record of a person's speech and writing i.e. some of their outputs.

Could a software be devised which gave the equivalent of a 3rd party observation on your choices? Could you even identify choices from the data?

That's just another example of real research with nuanced conclusions being transformed into pseudo-science for pop-psychology and business management books. Almost every book I read in this area I'm at least familiar with some of the research, or end up reading some of it as I go along. It seems like all books in this class of non-fiction best sellers overstates their conclusions in almost exactly the same way.

Same book, but obviously we have different "triggers". Like 'freshhawk', I found Kahneman's book to be astonishingly full of faulty logic and difficult to finish, but I have no problem reading Gladwell. Here's an HN thread from a a few years ago with my summary of Thinking, Fast and Slow: https://news.ycombinator.com/item?id=3431815

Perhaps it's that Kahneman is in the "uncanny valley" between science and popular writing. Since Gladwell is (for me) safely on the popular side, I find him engaging rather than misleading. But I'm pushed away by Kahneman's apparent overconfidence in places that I'm certain he's wrong, including such examples as Jason gives.

Ah, I found a pdf of the book and did some searches to jog my memory. It was probably the section where he describes going to walks with one of his associates to discuss ideas about decision making, bias and rationality and "examining our own intuitions" to gauge how closely they fit rational actor economic models or their proposed alternate models.

I find this whole "well I examined my intuitions about X so that is some evidence of this model of human decision making about X" that is immensely popular in philosophy (especially meta-ethics) and economics to just be embarrassing garbage. So that definitely hit a very strong bias I have. And it seemed completely out of place in a book concerned with human irrationality.